This multi-center, retrospective cohort study was conducted at the Hebei Cangzhou Integrated Traditional Chinese and Western Medicine Hospital and the Tianjin Cancer Hospital. Consecutive patients with newly diagnosed EGFR-mutant NSCLC and synchronous brain metastases who initiated first-line therapy between January 2019 and July 2023 were screened for inclusion. The inclusion criteria were as follows: (1) histological or cytological confirmation of lung adenocarcinoma; (2) diagnosis of brain metastases supported by contrast-enhanced magnetic resonance imaging (MRI); (3) confirmation of a sensitizing EGFR mutation (exon 19 deletion or L858R point mutation) via next-generation sequencing or polymerase chain reaction-based testing of tumor tissue or plasma circulating tumor DNA; (4) treatment with a first-line third-generation EGFR-TKI (osimertinib, aumolertinib, or furmonertinib) with or without brain-directed radiotherapy; (5) age ≥ 18 years; and (6) availability of complete clinical and follow-up data. Key exclusion criteria included: (1) Eastern Cooperative Oncology Group (ECOG) performance status (PS) ≥ 3 at the initiation of first-line therapy; (2) diagnosis of leptomeningeal metastasis at baseline; (3) active pregnancy or lactation; (4) history of other active malignancies within the past 5 years; and (5) prior radiotherapy to the brain before the initiation of first-line TKI therapy. Based on the initial treatment strategy, eligible patients were classified into three groups for analysis: the Early Combined Therapy (ECT) group, the Salvage Radiotherapy (SRT) group, and the TKI Monotherapy (TKI-Mono) group. The specific definitions for group assignment are detailed in the subsequent section.

Patients were categorized into three groups based on the timing and intent of brain radiotherapy relative to first-line TKI initiation. Early Combined Therapy (ECT) Group: Patients who received brain radiotherapy within 90 days of starting first-line TKI, in the absence of intracranial progression. This represented a planned, upfront combined-modality strategy. Salvage Radiotherapy (SRT) Group: Patients who began first-line TKI alone and received brain radiotherapy only after radiographic confirmation of intracranial progression. This represented a deferred, as-needed salvage strategy. TKI Monotherapy (TKI-Mono) Group: Patients treated with first-line TKI alone, who did not receive any form of brain radiotherapy during the entire study follow-up period. Systemic Therapy: All patients received a standard dose of a third-generation EGFR-TKI: osimertinib (80 mg orally once daily), aumolertinib (110 mg orally once daily), or furmonertinib (80 mg orally once daily). Treatment continued until disease progression, unacceptable toxicity, or patient withdrawal. Radiotherapy Technique: All radiotherapy was planned using a dedicated treatment planning system (TPS) with CT simulation and co-registration of contrast-enhanced brain MRI for target delineation. For patients with 1–3 brain metastases (oligometastatic disease), stereotactic radiosurgery/radiotherapy (SRS/SRT) was employed. The planning target volume (PTV) was generated by adding a 3 mm margin to the gross tumor volume (GTV). The prescription isodose line was typically 70%-90%. Prescribed doses ranged from 30 to 45 Gy in 5 to 10 fractions (median 5 fractions), with the specific regimen chosen based on tumor size, location, and physician judgment, aiming to cover ≥95% of the PTV. For patients with >3 brain metastases, hippocampal-avoidance whole-brain radiotherapy (HA-WBRT) with a simultaneous integrated boost (SIB) to individual metastases was used. The whole brain (excluding hippocampi) was treated to 30 Gy in 10 fractions (3 Gy/fraction). Individual metastases received a simultaneous integrated boost; the boost dose prescribed to the GTV of metastases ranged from 10 to 20 Gy in 5–10 fractions (2 Gy/fraction). Delineation, Constraints, and Quality Assurance: Hippocampal contouring followed the RTOG 0933 protocol guidelines (20). A comprehensive set of dosimetric and clinical parameters was collected for all radiotherapy courses to address potential variability. This included: for SRS/SRT – prescription dose, number of fractions, biologically effective dose (BED10, assuming α/β=10), prescription isodose line, and conformity index; for HA-WBRT – whole brain dose, boost dose per metastasis, and mean/maximum dose to the hippocampi. Organs-at-risk (OAR) constraints were strictly applied: lens maximum dose (Dmax) <7 Gy; optic nerves/chiasm and brainstem Dmax <54 Gy; pituitary gland mean dose (Dmean) <45 Gy; and hippocampal Dmax maintained between 9–16 Gy as per protocol, with a goal of Dmean ≤10 Gy. The anatomic location of each brain metastasis (supratentorial vs. infratentorial) was also documented. All treatment plans underwent review and approval by a multidisciplinary tumor board including a specialized neuro-radiation oncologist. These detailed radiotherapy parameters are summarized for the ECT and SRT groups in Supplementary Table 1.

Efficacy Endpoints: The primary endpoint of this study was intracranial progression-free survival (iPFS), defined as the time from the initiation of first-line TKI therapy to intracranial disease progression as per Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1 (21), or death from any cause, whichever occurred first. Key secondary endpoints included: overall survival (OS), measured from treatment initiation to death from any cause; progression-free survival (PFS), defined as the time to systemic or intracranial progression or any-cause death; intracranial objective response rate (iORR), calculated as the proportion of patients achieving a complete response (CR) or partial response (PR) as the best intracranial response; and intracranial disease control rate (iDCR), defined as the proportion of patients with a best response of CR, PR, or stable disease (SD) lasting for at least 6 weeks. Radiologic and Response Assessment: Tumor response was evaluated using contrast-enhanced computed tomography (CT) of the chest and abdomen and contrast-enhanced MRI of the brain. Baseline imaging was performed within 4 weeks before treatment initiation. Follow-up assessments were conducted every 8–12 weeks thereafter, or earlier if clinically indicated. Intracranial and extracranial responses were evaluated separately according to RECIST 1.1 criteria by two independent radiologists, with any discrepancies resolved by consensus. Safety and Toxicity Assessment: Treatment-related adverse events (AEs) were graded throughout the treatment and follow-up period according to the National Cancer Institute’s Common Terminology Criteria for Adverse Events (CTCAE) version 5.0 (22). Specific attention was paid to neurologic toxicities. Symptomatic radiation necrosis was diagnosed based on clinical symptoms correlated with characteristic imaging findings and requiring medical intervention. Neurocognitive function was assessed using the Mini-Mental State Examination (MMSE) at baseline (23), and at 1, 3, and 6 months after the completion of radiotherapy for patients in the ECT and SRT groups. A decline of ≥3 points from baseline was considered clinically significant cognitive deterioration. Late radiographic changes were evaluated using the Fazekas scale for white matter lesions on T2/FLAIR MRI sequences.

Patients were followed from the initiation of first-line therapy until death, loss to follow-up, or the study cutoff date of January 31, 2025. Follow-up was conducted through regular clinical visits, inpatient records, and telephone interviews. Clinical, imaging, and laboratory data were collected at each visit. Survival status was ascertained from hospital records or, when necessary, through direct contact with patients or their families. The median follow-up duration was calculated using the reverse Kaplan-Meier estimator.

This study involved two tertiary care institutions. While treatment decisions (ECT vs. SRT) were made historically by local multidisciplinary teams, we implemented a retrospective harmonization process. All cases were re-evaluated against the pre-defined study criteria to ensure consistent group assignment. Both centers utilized modern radiotherapy techniques (SRS/SRT and HA-WBRT with hippocampal avoidance) and followed national guidelines for systemic therapy. To assess the potential for center-specific bias, we compared the distribution of key baseline characteristics and the selection of radiotherapy technique between centers. No major systematic differences were identified that would compromise the validity of pooling data for the primary comparative analysis.

Statistical analyses were performed using SPSS (version 26.0) and R (version 4.3.1). Categorical variables were compared with the Chi-square or Fisher’s exact test, and continuous variables with the Mann-Whitney U or Kruskal-Wallis test. Survival curves were generated using the Kaplan-Meier method and compared with the log-rank test. Hazard ratios (HR) and 95% confidence intervals (CI) were derived from Cox proportional hazards models, both univariable and multivariable. A two-sided P value < 0.05 was considered statistically significant.

A total of 193 patients with treatment-naïve EGFR-mutant NSCLC and brain metastases were included in this retrospective analysis. Based on the initial treatment strategy, patients were stratified into three cohorts: the ECT group (n=83), the SRT group (n=83), and the TKI-Mono (n=27). The baseline demographic and clinical characteristics of the three groups are summarized in Table 1. No statistically significant differences were observed among the groups regarding sex, age, smoking history, KPS, comorbidities, EGFR mutation subtype, presence of extracranial metastases, number and size of brain metastases, presence of baseline intracranial symptoms, or the choice of first-line third-generation EGFR-TKI (all P > 0.05). The distribution of radiation technique between the ECT and SRT groups showed no significant difference (P = 0.080). The overall comparability of baseline features supports the validity of subsequent comparative efficacy analyses.

Primary Efficacy Analysis ECT versus SRT: The primary survival comparison between the ECT and SRT strategies demonstrated superior efficacy for the ECT approach across key endpoints. The median OS was significantly longer in the ECT group than in the SRT group (37.5 months vs. 31.8 months; HR 0.637, 95% CI 0.465–0.871; log-rank P = 0.003) (Figure 1). The benefit in intracranial disease control was particularly pronounced. The median iPFS was 22.4 months in the ECT group compared to 15.7 months in the SRT group (HR 0.628, 95% CI 0.459–0.858; P = 0.002) (Figure 2). A consistent, though less pronounced, trend favoring the ECT strategy was observed for systemic PFS, with median PFS of 17.7 months versus 15.0 months (HR 0.748, 95% CI 0.549–1.020; P = 0.058) (Figure 3). Sensitivity Analysis: ECT versus Initial Non-Combined Therapy: To assess the robustness of the primary findings, a sensitivity analysis was performed by combining the SRT group with the TKI-Mono group (n=27; median OS 26.5 months, median iPFS 13.6 months, median PFS 13.2 months) into an “Initial Non-Combined Therapy” cohort (total n=110). The survival advantage of the ECT strategy remained significant and was numerically enhanced in this comparison. The ECT group maintained a superior median OS compared to the combined cohort (37.5 months vs. 28.7 months; HR 0.589, 95% CI 0.442–0.784; P = 0.000) (Figure 4). Similarly, the iPFS benefit persisted (median 22.4 months vs. 15.2 months; HR 0.590, 95% CI 0.444–0.784; P = 0.000) (Figure 5). The PFS difference also reached statistical significance in this analysis (median 17.7 months vs. 14.3 months; HR 0.706, 95% CI 0.531–0.937; P = 0.015) (Figure 6).

The iORR and iDCR following treatment are detailed in Table 2. The iORR was significantly higher in the ECT group compared to the SRT group (81.9% vs. 66.3%, P = 0.021). In the TKI-Mono group, the iORR was 59.3%. The overall difference in iORR across all three groups was also statistically significant (P = 0.022). The iDCR was high in all groups, reaching 100% in the ECT group, 97.6% in the SRT group, and 96.3% in the TKI-Mono group. No significant difference in iDCR was observed between the ECT and SRT groups (P = 0.497) or across the three groups collectively (P = 0.363).

To identify independent prognostic factors, multivariate Cox proportional hazards analyses were performed for OS and iPFS, adjusting for key baseline and treatment-related variables, including radiotherapy technique (SRS/SRT vs. HA-WBRT). The results are presented in Tables 3, 4. In the multivariate model for OS (Table 3), the ECT strategy remained an independent factor associated with significantly improved survival after adjustment for other variables (adjusted HR 0.602, P = 0.009). The only other independent prognostic factor was a higher number of brain metastases (>3), which was associated with worse OS (adjusted HR 1.852, P<0.001). The analysis for iPFS yielded consistent and stronger results (Table 4). The ECT strategy was independently associated with a markedly reduced risk of intracranial progression (adjusted HR 0.523, P<0.001). Additionally, the use of stereotactic radiotherapy (vs. WBRT) was associated with superior intracranial control (adjusted HR 0.621, P = 0.009). As expected, a higher number of brain metastases was the strongest independent adverse prognostic factor for iPFS (adjusted HR 2.215, P<0.001). The type of EGFR-TKI used, age, EGFR mutation subtype, and presence of extracranial metastases were not independent prognostic factors for either endpoint in the adjusted models.

Subgroup analyses consistently favored the ECT strategy across most patient categories for both OS and iPFS (Figures 7, 8). The survival benefit was particularly pronounced in key clinical subgroups. Patients with EGFR exon 19 deletion derived greater benefit in both OS (HR 0.52, P = 0.005) and iPFS (HR 0.45, P<0.001) compared to those with the L858R mutation. Similarly, the advantage of the ECT strategy was most significant among patients with 1–3 brain metastases for OS (HR 0.50, P = 0.002) and iPFS (HR 0.42, P<0.001). A consistent treatment effect was observed across other subgroups, including those defined by age, sex, KPS, and the presence of extracranial metastases. To visually complement the subgroup analyses and further assess the impact of intracranial tumor burden, we performed additional stratified analyses based on the number of brain metastases (1–3 vs. >3) and the largest brain metastasis diameter (<1 cm vs. ≥1 cm). Kaplan-Meier curves comparing the three treatment strategies (ECT, SRT, and TKI-Mono) within each stratum are presented in Supplementary Figures 1-8. Analysis by Number of Brain Metastases: Among patients with 1–3 brain metastases, the ECT group demonstrated significantly longer iPFS (median 25.1 months) and OS (median 38.9 months) compared to both the SRT (iPFS: 18.7 months; OS: 32.6 months) and TKI-Mono groups (iPFS: 14.7 months; OS: 27.2 months; P = 0.001 for both iPFS and OS) (Supplementary Figures 1, 2). In patients with >3 brain metastases, the ECT strategy was also associated with significantly longer iPFS (median 15.8 vs. 13.8 vs. 10.1 months, P = 0.027) and a non-significant trend toward longer OS (median 28.1 vs. 23.5 vs. 19.4 months, P = 0.071) compared to SRT and TKI-Mono, respectively (Supplementary Figures 3, 4). Analysis by Largest Brain Metastasis Diameter: Similarly, in the subgroup with the largest brain metastasis diameter <1 cm, the ECT group had significantly improved iPFS (median 22.3 months) and OS (median 37.5 months) compared to the SRT (iPFS: 14.6 months; OS: 27.8 months) and TKI-Mono groups (iPFS: 11.6 months; OS: 21.3 months; P = 0.014 and P = 0.042, respectively) (Supplementary Figures 5, 6). For patients with lesions ≥1 cm, ECT was associated with significantly longer iPFS (median 23.4 vs. 16.8 vs. 13.6 months, P = 0.006) and a trend toward longer OS (median 37.9 vs. 32.1 vs. 26.5 months, P = 0.090) (Supplementary Figures 7, 8).

Detailed radiotherapy parameters for the ECT and SRT groups are summarized in Supplementary Table 1, demonstrating overall comparability in prescription doses, fractionation schemes, and adherence to organs-at-risk constraints between the two strategy groups. Exploratory Analysis of Radiotherapy Technique: To address the potential impact of radiotherapy modality, we first compared outcomes between patients who received SRS/SRT (n=101) and those who received HA-WBRT (n=65), irrespective of treatment timing. Patients treated with SRS/SRT had significantly longer iPFS than those receiving HA-WBRT (median 21.4 vs. 16.4 months; HR 0.716, 95% CI 0.516–0.993; P = 0.033) (Supplementary Figure 9). However, there was no significant difference in OS between the technique groups (median 34.5 vs. 32.1 months; HR 0.885, 95% CI 0.645–1.215; P = 0.435) (Supplementary Figure 10) or in systemic PFS (median 17.0 vs. 14.0 months; HR 0.834, 95% CI 0.606–1.148; P = 0.246) (Supplementary Figure 11). Effect of Treatment Strategy Within Radiotherapy Subgroups: We next examined whether the benefit of the ECT strategy was consistent across the different radiotherapy modalities through pre-specified subgroup analysis (Figures 7, 8). The survival advantage of ECT over SRT was significant within the subgroup of patients treated with SRS/SRT for both OS (HR 0.55, 95% CI 0.360–0.840; P = 0.006) and iPFS (HR 0.47, 95% CI 0.310–0.720; P<0.001). In the HA-WBRT subgroup, a trend favoring ECT was observed, but it did not reach statistical significance for OS (HR 0.71, 95% CI 0.430–1.170; P = 0.177) or iPFS (HR 0.66, 95% CI 0.400–1.090; P = 0.103).

The safety profile of the ECT strategy was comparable to that of the SRT strategy. As detailed in Table 5, the incidence of grade 3 or higher AEs was similar between the ECT and SRT groups (33.7% vs. 30.1%, P = 0.617). The spectrum of AEs was consistent with the known profiles of third-generation EGFR-TKIs and cranial radiotherapy. No significant between-group differences were observed in the rates of specific TKI-related toxicities (e.g., rash, diarrhea), acute radiotherapy-related events (e.g., headache, nausea), or hematologic toxicities (all P>0.05). Of critical importance, the incidence of key neurologic toxicities was low and did not differ between the two strategies. Symptomatic radiation necrosis occurred in 2.4% of ECT and 1.2% of SRT patients (P = 1.000). Similarly, the rates of symptomatic cerebral edema requiring steroid intervention (3.6% vs. 4.8%, P = 1.000) and objective cognitive decline (4.8% vs. 6.0%, P = 1.000) were comparable.